This document outlines the security measures implemented in the Root Ventures Claude Apply Skill and known limitations.

• JSON Injection Protection: All user inputs are passed to jq using --arg flags, which treats them as literal strings and properly escapes special characters
• No eval/exec: User input is never executed as code
• Validation: Required fields (name, email) are validated before processing

• mktemp: Uses mktemp to create secure temporary files with unpredictable names
• Cleanup: Trap ensures temporary files are deleted even on script failure
• Permissions: Temporary files inherit secure permissions from mktemp

• Encrypted Transport: All downloads use HTTPS to prevent man-in-the-middle attacks
• GitHub Trust: Files downloaded from GitHub's trusted CDN

• set -e: Script exits immediately on errors to prevent cascading failures
• Graceful Degradation: Clear error messages guide users when issues occur

• Read-Only Access: Script only reads from user's home directory for skill files
• Write-Only Webhook: Attio webhook can only create records, not read existing data
• No Persistent State: No data stored locally after submission

Status: Accepted Risk

Description: The Attio webhook URL is public in the GitHub repository.

Why It's Acceptable:

• Webhook is write-only (can't read existing data)
• Attio provides rate limiting
• Intended for public job applications
• No sensitive data exposure

Mitigation: Monitor Attio for spam applications, use webhook secret tokens if needed

Status: Accepted Risk

Description: Downloaded files (skill.json, prompt.txt, apply.sh) are not cryptographically verified.

Why It's Acceptable:

• HTTPS provides transport security
• GitHub provides platform security
• Low-value target for attackers
• User can manually inspect files before running

If You Want Extra Security:

# Manually verify downloaded files
cd ~/.claude/skills/root-ventures-apply
cat apply.sh  # Review the script
cat prompt.txt  # Review the prompt

Description: Assumes user running the skill has appropriate permissions on their system.

Mitigation: Script only modifies user's home directory (~/.claude/skills/)

1. User Input → Claude CLI → apply.sh arguments
2. apply.sh → jq (sanitizes) → JSON payload
3. JSON payload → HTTPS POST → Attio webhook
4. Attio → Stores in your workspace

1. Review the code: Check the GitHub repository
2. Verify URL: Ensure you're downloading from github.com/rootvc/claude-apply-skill
3. Check HTTPS: Browser should show secure connection

1. Inspect files: Review what was installed in ~/.claude/skills/root-ventures-apply/
2. Understand permissions: The skill can read/write in your home directory
3. Monitor submissions: Check your Attio workspace for unexpected applications

If you discover a security vulnerability:

DO NOT open a public GitHub issue

DO email: security@root.vc with:

• Description of the vulnerability
• Steps to reproduce
• Potential impact
• Suggested fix (if any)

We'll respond within 48 hours.

We monitor for:

• Dependency vulnerabilities (jq, curl, bash)
• GitHub security advisories
• Attio API changes

Updates will be pushed to the main branch and noted in commits.

• Input sanitization (jq --arg)
• Secure temporary files (mktemp)
• HTTPS for downloads
• Error handling (set -e)
• No arbitrary code execution
• Minimal file system access
• Clear data flow
• Documentation of risks
• Cryptographic verification (future enhancement)
• Webhook secret validation (future enhancement)

1. Checksum Verification

   • Add SHA256 checksums for downloaded files
   • Verify before execution

2. Webhook Secrets

   • Add HMAC signature to webhook requests
   • Verify requests are from legitimate skill users

3. Rate Limiting

   • Client-side rate limiting to prevent abuse
   • Cooldown between submissions

4. GPG Signatures

   • Sign releases with GPG
   • Verify signatures during installation

• No PII Storage: Application data goes directly to Attio, nothing stored locally
• User Consent: Users explicitly provide information through conversation
• Data Minimization: Only collects necessary fields

• Users provide data voluntarily
• Data goes to Attio (processor)
• Root Ventures (controller) has privacy policy
• Users can request deletion from Attio

Last Updated: January 27, 2026 Security Contact: security@root.vc